by Noah Millman
December 23, 2008
As readers of this blog may recall, I’ve spent the last several years working in an industry now credibly blamed for bringing on what is almost certainly the worst economic crisis since the Great Depression. Not only have I been working on Wall Street, but I work in structured finance. While I wasn’t at the epicenter of the crisis – I didn’t put together mortgage securitizations, and I don’t work for one of the big firms – I was certainly using some of the same financial technology, and trading the same products, as the businesses that were at the epicenter.
Given my eye-of-the-storm view of the matter, I thought it would be of interest to relate two stories from my long career in structured finance, one that may help explain why, if you asked me in 2004 or 2005, I would have staunchly defended structured finance technology as having real social benefit and why, by a couple of years later, it was clear to anyone looking honestly at the business that something had gone very wrong. I’ve chosen two stories that, as it happens, do not relate to mortgage securitization, for two reasons: first, because my world was primarily centered on corporate structured finance, so that’s what I know best, but, second, because I want to make it clear that both the promise and the peril of this financial technology extend well beyond the best-reported area of mortgage finance.
First, the promise.
Most of what I’ve worked on over the years is structuring and issuing synthetic corporate CDOs, and actively adjusting the composition of those asset pools over time. Our firm closed its first synthetic CDO in late 2001. To refresh everyone’s memory, late 2001 saw the terrorist attacks on New York and Washington, the collapse of Enron, and, following shortly thereafter in early 2002, the collapse of Worldcom. Enron and Worldcom were, at the time, the two biggest corporate frauds in history, and their collapse had far-reaching consequences.
Enron and Worldcom had invested huge amounts of mostly borrowed money in, respectively, a variety of commodity trading operations and broadband telecommunications infrastructure. Both firms appeared to be profitable, and numerous other firms – energy-generation utilities in the first case, telecom providers in the second – borrowed lots of money to invest in competing platforms, so as not to be left behind. When it became clear that both firms were frauds at heart, and that these activities were not profitable, it didn’t take long for people to figure out that there were a whole lot of other firms out there that were now in trouble. The market had already priced in a slowdown in telecom, for example, but it thought this slowdown was from a base of real profitability. If there had been no profit when times were good, then things were going to be much worse than originally anticipated now that times were bad.
Everyone knows what happened to the stock market in 2002. The corporate debt market situation was not as well-reported because it was of less interest to individual investors. But it was of enormous importance to the future of these companies, and to the world economy. Many of these companies had huge debt loads and substantial near-term refinancing needs. And it wasn’t clear they were going to get that refinancing, as investors were not eager to buy bonds from companies that might be the next Enron or Worldcom, and banks were not eager to lend more when they were increasing their provisions against their existing holdings. There was a real danger of a credit crunch, and a catastrophic series of bankruptcies that would cripple the banking sector for years.
Everyone also knows what the Fed did to respond to this threat: it dropped interest rates to what were once considered extremely low levels (the same thing the Fed did after the S&L debacle). Since banks borrow short via deposits and lend long, dropping the front end of the curve straightforwardly increases banks’ profitability which should (if banks are well-capitalized) make them more willing to risk capital by lending. And lend they did: most of the companies in question got new financing and, over the next two years, corporate America dramatically improved its balance sheets, paying down and/or extending out its debt, setting itself up for several years of extraordinary profit growth.
But, once again, what’s less reported is the role that structured finance played in facilitating bank lending. And it played a very important role. One reason banks were willing to lend again was that the credit derivatives marketplace made it possible for them to price and hedge their credit risk. Prior to the development of credit derivatives, it was extremely difficult for banks to get a clear picture of the price of credit. There was no market for corporate loans, and the corporate bond market is a clubby place where transparent pricing is available only for the most liquid issuers. But by 2002, the credit derivatives marketplace had matured to the point that banks could use it as one input into their decisionmaking process about when to lend.
And they could hedge their risk as well. Let’s say you were a utility that needed to roll over $1 billion in debt coming due in six months. The syndicate that lent you the money initially is ten banks, so that’s $100 million per bank on average. None of the banks want to take that much risk. At their previous level of security, they’d want to cut back to $25 million each; if they can get additional security, they’d consider going to $50 million each. How do you increase lending capacity? Well, if a bank can buy a credit default swap on $50 million, then it can lend $100 million, and hedge itself back to $50 million. And who would write that default swap? Maybe a hedge fund making a bet that credit would outperform equity (so-called “capital structure arbitrage”) and was therefore writing credit-default-swaps and shorting stock as a hedge. Maybe an insurance company taking advantage of wide spreads to take additional corporate credit exposure in its investment portfolio. Maybe a foreign bank that isn’t part of the lending syndicate for the utility looking to diversify geographically.
Or maybe the protection seller was a structured finance vehicle. This could be a vehicle set up by the bank itself – a balance-sheet CDO. This was an increasingly popular technology early in the decade, and it enabled banks to reduce concentrations to a wide variety of corporate credit risks all in one shot. Basically, the bank would take its portfolio of loans, and transfer a portion of these loans to a special-purpose vehicle via a credit default swap. Then the vehicle would issue various classes of debt, backed by government bonds and by the cashflow from this default swap, rated based on their relative level of seniority. The ratings agency process for rating this debt was based, as always with the agencies, on their historical data on corporate defaults and recoveries on defaulted assets in workout, and, given that corporate default data goes back to the 1920s, they’ve got a pretty decent data set to work off – moreover, this is the same data set that corporate bond investors such as insurance companies have long used in their own investment process, so market confidence in the ratings methodology was pretty high. If a bank set up such a vehicle, it could significantly expands its capacity to lend – it was effectively creating a synthetic syndicate for every loan it needed to hedge, a syndicate consisting not of other banks but of a much wider array of other investors, each investing at its preferred level of risk.
And apart from bank balance sheet CDOs, there were also “arbitrage” CDOs – investment vehicles structured around CDO technology. Equity investors in such vehicles were basically borrowing for term from the various classes of debt investor to purchase a portfolio of corporate risk, and could change the composition of that portfolio (subject to a variety of ratings agency restrictions) to capitalize on market moves of various kinds. By 2002, this CDO market had already become synthetic, which meant that a variety of investment banks could create such structures on the fly designed around the precise risks investors wished to access (in terms of the underlying portfolio and the degree of leverage in the structure). Again, because of the robustness and familiarity of the data set used by the agencies, market confidence in the ratings process was fairly high. Arbitrage vehicles of this sort made a material contribution to the liquidity of the corporate credit default swap market in the early 2000s, and thereby made it possible for banks to hedge their credit exposures to corporations to whom they extended credit, which in turn made it possible for them to expand their lending activities at precisely the moment when they were most nervous about doing so, and when the economy most needed them to do so.
Coming out of the experience of 2002, I can fairly say that I was a believer in the social value of credit derivatives and structured finance as part of the great machine of capitalism. I had seen this market and this technology play an important part in speeding the recovery from a fairly serious corporate borrowing binge that started with the internet bubble and ended with the collapse of Enron and Worldcom. Not the leading part – the Fed’s rate cuts were clearly much more important – but an important part nonetheless. Credit derivatives and structured finance were supposed to increase market transparency and liquidity, facilitate the more efficient pricing and distribution of credit risk, and hence smooth out the credit cycle. And that’s exactly what they appeared to have done.
Then, the peril.
Fast-forward to 2006. I could easily spend 10,000 words explaining all the ways in which the credit-derivatives and structured-products markets had grown, and changed. By this time, structured finance technology had facilitated a truly insane inflation of the housing market, and the extension of credit on absurd terms to borrowers who nobody could rationally expect to be able to pay. And credit derivatives made it possible for companies like AIG to leverage their AAA credit ratings to acquire enormous nominal amounts of purportedly loss-remote credit risk and leverage this risk hundreds of times without any regulatory body being aware of their activities, much less able to intervene to control them. Much of this has been well-reported in a variety of venues. I want to tell one anecdote from the corporate structured finance market (the market I know best) as a means of debunking what I think is a rather self-justifying explanations for how things went so wrong in the mortgage market.
It has been argued that, in a nutshell, the reason things went so wrong in the structured finance market is that “nobody” thought that the national housing market could suffer a meaningful year-on-year price decline. Local markets could decline, yes, and that national market could stagnate for a bit, as it had in past recessions. But a significant national decline in housing prices was inconceivable, and hence was not factored in to the ratings-agency models for mortgage securitizations. Rather, the agencies assumed historic default correlations between different geographies were predictive of future correlations, as well as that historic default rates were predictive of future default rates.
The argument against this is that the agencies should have known that the structured finance market they created by rating all this paper was, itself, changing the complexion of the underlying mortgages, and that this could change both default rates and default correlations (as, indeed, occurred). They should have factored this “reflexivity” (to use George Soros’ term) into their models, and made their ratings tougher. And there’s a great deal of truth to this, but I think this objection gives the agencies too much credit. After all, if all they got wrong was not paying attention to the tails of the distribution – the “black swan” problem – then they made the kind of error that happens all the time in finance; indeed, the kind of error that routinely creates bubbles. The thing about black swans is that you have no idea how likely they are, which makes it extremely difficult to build them into your models. There are always Cassandras out there saying the sky is about to fall, and eventually they will be right, but if you listen to them all the time you’ll never get out of bed.
The agencies got so much more wrong than that (for example, they didn’t rate the quality of servicers or originators of loans, which you would think would be important in assessing the credit quality of the mortgages in a securitization). But the thing they got most wrong was ignoring the quality of their own data set. The sub-prime mortgage market was still young when the agencies began rating securitizations dominated by such mortgages. And the structured finance market was in its infancy when the agencies began rating CDOs backed by asset-backed securities backed in turn by sub-prime mortgages. There is simply no way that the agencies had an adequate data set to justify rating these deals at all. And I can best illustrate the agencies’ disregard for the quality of their data sets by reference to a product unrelated to mortgages: the constant-proportion debt obligation, or CPDO.
This obscure little product was launched in late 2006 to great fanfare. And it was, indeed, an ingenious little fraud of a product. Here’s how it worked.
Some clever structurer noticed that, historically, investment-grade companies don’t default very often. Rather, they deteriorate for a while first, get downgraded to junk status, and eventually they default. That’s why the short-term ratings for relatively low-rated investment-grade companies may be reasonably high: even BBB companies are generally good for the next 90 days; if they weren’t, they wouldn’t be rated BBB. And this suggested the possibility of a trading strategy.
What if you bought a portfolio of investment-grade corporate debt and, every six months, purged it of the bonds that were downgraded to junk, replacing these with new investment-grade bonds. Obviously, you’d expect the downgraded bonds to underperform the remainder of the pool, so you’d have losses you need to make up, so assume you also sell out of a handful of bonds that have done well, and replace these with higher-yielding bonds that are still investment-grade, thus keeping your yield relatively constant. You’d still expect some losses if you wanted to keep your average rating relatively constant, though. So you need some excess yield to make up for these losses.
You find that yield by leveraging the portfolio. After all, it’s an investment-grade portfolio, very unlikely to default in the short term. You can borrow very cheaply short-term, invest in longer-term bonds, and earn the spread differential. If the bonds go against you, that’s OK, because you’re going to hold them to maturity and you’ll always be able to roll your short-term borrowing. And, if you can get a high enough degree of leverage, the excess in current yield from the differential between where you borrow and the yield on your portfolio should more than pay for the cost of rolling out of your losers every six months. And if you do that successfully, you’ve got a trading strategy that never loses principal, but has a surprisingly high expected yield. Sound good?
Well, it sounded great to the ratings agencies, who blessed this strategy by giving it a AAA rating.
How did they justify that AAA rating? By looking at the historic cost of rolling credit derivatives on indices of investment-grade corporate issuers, which generally have a high-BBB rating. These had been around for about three years when the first CPDOs were rated, and the roll had never cost more than 3 basis points. Factoring in that cost, at a leverage of 15-to-1, and using historic 6-month default rates for the portfolio (since the index would be rolled every six months), the proposed trading strategy would never lose money. Hence a AAA rating.
Let me reiterate that, just to drive the point home. The ratings agencies said: you can take a BBB-rated index, leverage it 15-to-1, and follow an entirely automatic trading strategy (no trader discretion, no forecasting of defaults or anything, just a formula-driven adjustment to the leverage ratio and an automatic roll of the index), and the result is rated AAA.
Needless to say, this worked out really well for all concerned. But that’s not really the point. The point is: the notion that you could grant a AAA based on a trading strategy for which there was at best three-years of data (three years that encompassed not a single recession, I’ll note) is mind-boggling. And, worse than that, nobody at the agencies apparently stood up and said, “wait a second: how can you turn a BBB into a AAA by leveraging it 15-to-1? That’s impossible!” Which, of course, it is.
I want to be very clear about something: we’re not talking about a CDO where the AAA investor is providing leverage, and there are subordinate investors below who bear the first risks of loss. This was a trading strategy; the investor in the AAA CPDO had first-loss risk with respect to a BBB portfolio. The trading strategy was just supposed to generate enough returns to create “virtual” subordination to justify the AAA.
When I first heard about this product, I thought: whichever agencies rated this thing have lost their minds. When people asked me whether it made sense as an investment, I said: it’s an outright fraud. You’re practically guaranteed to lose money. I never bought or sold one of these things myself, and neither did anyone else in our group. But the existence of such a ridiculous product should have been a wake-up call about just how divorced from reality the agencies were. And if they were out to lunch on something as straightforwardly absurd as the CPDO, how out to lunch were they on other products, ones that were far more significant to the markets and the economy, where the absurdity of their assumptions was less-obvious?
How did a market that, I thought, had really helped capitalism work in 2002 become the great destroyer of capitalism of the last two years? There were a lot of contributors to the catastrophe, but one indispensable one is that the ratings agencies monetized their sterling reputations in an extraordinary fashion, and nobody in regulatory apparatus of government saw that this was happening, and what it might portend. The success of 2002 depended on market confidence in the ratings agency process: that’s what made investors willing to buy the notes issued by structured finance vehicles that issued the credit protection that made it possible for banks to hedge. Without that confidence, the market would never have developed. And by 2006, the agencies understood just how much that confidence was worth.
The ratings agencies have an enormous amount of power: pension funds and insurance companies invest according to their rules; under Basel II, bank capital ratios are substantially determined by how the agencies rate their portfolios of loans; and, of course, the entire “shadow banking system” created by providers of super-senior credit protection (monoline insurers, bank-sponsored asset-backed conduits, AIG Financial Products, etc) was only possible because of the ratings agencies. By 2006, the entire financial system was extraordinarily leveraged to the opinions of these government-blessed non-governmental independent agencies, and these agencies were monetizing their market position by trashing their process.
Now, of course, the agencies have radically reversed course. They have completely changed their models, of course. But they’ve also begun to “follow the market”, incorporating credit derivatives swap and equity market pricing into their ratings for companies. If in 2006 the market had, to an alarming degree, delegated its risk-management to the ratings agencies, now the ratings agencies have delegated a great deal of their ratings process to the market. And so the market has lost any reason for confidence in the agencies in both directions: they cannot be trusted when the market is strong to assess the downside risks the market is ignoring, and they cannot be trusted when the market is weak to assess a company’s financial condition independently of the market panic.
This collapse of confidence is having a material impact on the ability of the credit markets to find their footing. I don’t want to minimize the impact of other factors – rates held too low for too long, Wall Street greed, an Administration with an ideological aversion to regulation – but I want to emphasize this one because of its unique contribution to this particular bubble and because I don’t see an obvious solution to it. The old model is permanently broken, and we have not yet come up with an alternative. Right now, we’re in the low-trust environment that is the reality when the libertarian fantasy of eliminating the market-distorting regulators actually comes to pass. The market is inevitably focused on a short time horizon, fickle and volatile by its very nature. We want major financial institutions – banks, insurers, etc. – to look beyond the market to longer term risk metrics. Without the agencies as an independent arbiter of what these might be, the market is all we have left. Trust is a very hard thing to rebuild, and structural changes – having the agencies be government-sponsored, or paid by investors rather than issuers – are insufficient solutions (government-sponsored entities are also capable of seeking to monetize their position – look at Fannie Mae – and investor-sponsored agencies would be subject to the same pressures to facilitate the business that investors want to do).
We are still in a dark wood wandering, the right road lost.